Abstract: Medical image segmentation is currently performed by qualified medical personnel who, however, often must deal with large amounts of work, often in emergencies. Therefore, there is a need for automatic and reliable methods to analyze medical images, particularly magnetic resonance imaging, with high precision. There are several methods in the literature that are mainly based on the use of 3D atlases or convolutional neural networks, subject to several disadvantages. The present research work focuses on the development of an alternative and innovative method for the segmentation of medical images using a bilateral filter, a genetic programming algorithm developed ad hoc, the CLAHE algorithm, and the Watershed segmentation algorithm, typically used for geo-topographic images. The results obtained by the following methodology allow us to conclude that the proposed approach provides comparable results with a smaller dataset of images compared to the most common neural networks used in the field. The integration of the genetic algorithm represents a highly innovative solution, improving the local quality of the contrast and reducing the noise of the images, and allowing for optimizing the Watershed segmentation.

Keywords: Segmentation, Medical Imaging, Watershed, CLAHE, Genetic Programming, Magnetic Resonance

Pages: 169-180